On this page

Tigris is a globally distributed S3-compatible object storage service that
allows you to store and access any amount of data for a wide range of use cases.
Tigris automatically and intelligently distributes your data close to the users,
and removes the need for you to worry about the complexities of data
replication, and caching.

## How to use Tigris

Most teams adopt Tigris by configuring existing
[AWS S3 or Google Cloud Storage SDKs](/content/docs/sdks/s3/index.html) with Tigris
[access keys](/content/docs/iam/manage-access-key/index.html) and a
[Tigris endpoint](/content/docs/sdks/s3/aws-cli/#service-endpoints/index.html). In many cases,
applications can switch to Tigris with no code changes beyond configuration.

Tigris also offers [native Storage SDKs](/content/docs/sdks/tigris/index.html) that provide direct
access to Tigris-specific features like
[client uploads](/content/docs/sdks/tigris/client-uploads/index.html) and
[bucket forks and snapshots](/content/docs/buckets/snapshots-and-forks/index.html). For
AI-assisted development, the [Tigris MCP server](/content/docs/mcp/local/index.html) lets AI
coding agents interact with your Tigris buckets directly.

## What Tigris stores

Tigris stores objects—such as application assets, model weights, media files,
and ML artifacts—that are consumed by databases, analytics systems, vector
search engines, and AI pipelines. Tigris focuses on durable object storage and
does not currently provide databases or query engines. However, Tigris can
replace a traditional CDN for many use cases due to its automatic global
replication.

## When to choose Tigris

**You're building AI and data-intensive workloads that span clouds or**
**providers.** If you train on GPU neoclouds, run inference across multiple
providers, or want to avoid lock-in to a single cloud, Tigris gives you a
single, globally replicated object store. Data is stored and replicated close to
where it's accessed, reducing latency and eliminating egress fees when data
moves between clouds.

**You need a shared data layer for AI systems.** Tigris is commonly used to
store model weights, checkpoints, embedding files, feature data stored as
objects, and training datasets that are consumed by external training
frameworks, inference services, vector databases, and analytics systems. Because
Tigris does not charge egress fees, large datasets can be reused freely across
environments.

**You want isolated environments for agents and experiments.** Bucket forks let
AI agents, experiments, and evaluation runs work against isolated copies of the
same underlying data without collisions. Even very large datasets can be forked
instantly, making it practical to run parallel experiments at scale.

**You care about predictable costs for data-heavy workloads.** With no egress
fees, Tigris lets you move and reuse data without surprise bills. This is
especially valuable for AI training, batch processing, analytics, and media
workloads where data movement dominates cost.

**You're migrating from another S3-compatible provider.** Shadow buckets keep
your existing storage and Tigris synchronized, enabling zero-downtime migration.
Applications can switch over gradually, often with only configuration changes.

Typical use cases include:

- Storage for machine learning models and datasets
- Storage for real-time applications and AI-powered services
- Web content and media (images, video, static assets)
- Storage for IoT applications and globally distributed data ingestion
- Data analytics, big data, and batch processing
- Backups and archives

## What's next

Ready to dig deeper? Explore the full set of
[Tigris features](/content/docs/overview/features/index.html) — including global distribution,
snapshots and forks, zero egress fees, and more.

- [How to use Tigris](/content/docs/overview/#how-to-use-tigris/index.html)
- [What Tigris stores](/content/docs/overview/#what-tigris-stores/index.html)
- [When to choose Tigris](/content/docs/overview/#when-to-choose-tigris/index.html)
- [What's next](/content/docs/overview/#whats-next/index.html)